Principal Balances

Abstract

Principal balances are defined as a sequence of orthonormal balances which maximize successively the explained variance in a data set. Apparently, computing principal balances requires an exhaustive search along all possible sets of orthogonal balances. This is unaffordable for even a small number of parts. Three suboptimal, but feasible, alternatives are explored. The approach is illustrated using a data-set of geochemical composition of glacial sediments